Research On Forecasting Method Of Power Optical Fiber Line State

In the electric power communication system, in recent years, the power optical fiber has been widely used as a mainstream mean of communication because of its many advantages. As a key component of electric power communication system, the reliability of power optical fiber transmission network is an important guarantee to the safe production and efficient running of electric power communication system. Once a failure occurred in the power optical fiber line, the communication interrupt will be caused that will bring huge economic losses to the enterprises and users. Therefore, it has important practical significance to research the forecasting method of power optical fiber line state, which could forecast the future possible line failure according to the known state of power optical fiber line, and arrange the maintenance and management measure in advance to avoid the failure, so that to ensure the uninterrupted transmission of electric power communication system.In order to forecast the running state of power optical fiber line more accurately, this paper proposes a new forecasting method with ARIMA-SVM based on adaptive particle swarm optimization(APSO), the method is based on the forecasting of optical power trend that is a kind of measure index of power optical fiber line running state, so as to realize the forecasting of power optical fiber line state. Firstly, according to the nonlinear, time-varying and complexity characteristics of optical power data from the power optical fiber line, the wavelet transform is used to decompose and reconstruct the monitored optical power data to extract the random component and trend component in optical power data. Then, according to random component characteristics and trend component characteristics, various forecasting models are constructed to forecast, through comparing with the experiments, the ARIMA model and SVM model with the best forecasting effect are selected to forecast the random optical power data and the trend optical power data respectively. In order to further improve the model forecasting performance, this paper designs the APSO algorithm to optimize the parameters of SVM model. A new dynamic distance function and the adaptive inertia weight are defined in the APSO algorithm, which realizes the search speed of different particles changed adaptively, so as to improve the convergence speed and accuracy of the algorithm, and get more accurate model parameters. The availability of the APSO algorithm is verified by simulation experiments. Finally, the ARIMA model and the SVM model based on APSO are constructed and utilized to forecast the random and the trend optical power data respectively, and the forecasted results of above models were combined to complete the optical power future trend forecasting, so as to achieve the forecasting on power optical fiber line state. The experimental results show that, compared with other forecasting methods, the proposed hybrid forecasting method with ARIMA-SVM based on APSO could forecast the future trend of optical power in power optical fiber line more accurately, in order to realize the forecasting on power optical fiber line state, this method optimizes the model forecasting performance, compare to the forecasting method with single SVM model, the forecasting accuracy of the proposed method is improved by 12.6%; compare to the forecasting method with ARIMA-SVM hybrid model, the forecasting accuracy of the proposed method is improved by 53.8%; compare to the forecasting method with ARIMA-SVM based on LDW, the forecasting accuracy of the proposed method is improved by 26.5%.